Utilizing the universal approximation property of neural networks, we develop several novel approaches to neural network-based adaptive output feedback control of nonlinear systems, and illustrate these approaches for several flight control applications. In particular, we address the problem of non-affine systems and eliminate the fixed point assumption present in earlier work. All of the stability proofs are carried out in a form that eliminates an algebraic loop in the neural network implementation. An approximate input/output feedback linearizing controller is augmented with a neural network using input/output sequences of the uncertain system. These approaches permit adaptation to both parametric uncertainty and unmodeled dynamics. All physical systems also have control position and rate limits, which may either deteriorate performance or cause instability for a sufficiently high control bandwidth. Here we apply a method for protecting an adaptive process from the effects of input saturation and time delays, known as ``pseudo control hedging". This method was originally developed for the state feedback case, and we provide a stability analysis that extends its domain of applicability to the case of output feedback. The approach is illustrated by the design of a pitch-attitude flight control system for a linearized model of an R-50 experimental helicopter, and by the design of a pitch-rate control system for a 58-state model of a flexible aircraft consisting of rigid body dynamics coupled with actuator and flexible modes. A new approach to augmentation of an existing linear controller is introduced. It is especially useful when there is limited information concerning the plant model, and the existing controller. The approach is applied to the design of an adaptive autopilot for a guided munition. Design of a neural network adaptive control that ensures asymptotically stable tracking performance is also addressed.

Neural network-based adaptive output feedback approaches that augment a linear control design are described in this thesis, and emphasis is placed on their real-time implementation with flexible systems. Two different control architectures that are robust to parametric uncertainties and unmodelled dynamics are presented. The unmodelled effects can consist of minimum phase internal dynamics of the system together with external disturbance process. Within this context, adaptive compensation for external disturbances is addressed. In the first approach, internal model-following control, adaptive elements are designed using feedback inversion. The effect of an actuator limit is treated using control hedging, and the effect of other actuation nonlinearities, such as dead zone and backlash, is mitigated by a disturbance observer-based control design. The effectiveness of the approach is illustrated through simulation and experimental testing with a three-disk torsional system, which is subjected to control voltage limit and stiction. While the internal model-following control is limited to minimum phase systems, the second approach, external model-following control, does not involve feedback linearization and can be applied to non-minimum phase systems. The unstable zero dynamics are assumed to have been modelled in the design of the existing linear controller. The laboratory tests for this method include a three-disk torsional pendulum, an inverted pendulum, and a flexible-base robot manipulator. The external model-following control architecture is further extended in three ways. The first extension is an approach for control of multivariable nonlinear systems. The second extension is a decentralized adaptive control approach for large-scale interconnected systems. The third extension is to make use of an adaptive observer to augment a linear observer-based controller. In this extension, augmenting terms for the adaptive observer can be used to achieve adaptation in both the observer and the controller. Simulations to illustrate these approaches include an inverted pendulum with its cart serially attached to two carts (one unmodelled), three spring-coupled inverted pendulums, and an inverted pendulum with its initial condition in a range in which a linear controller is destabilizing.